[comp.lang.fortran] Assembly or HLL's on vector supercomputers

mccalpin@loligo.fsu.edu (John McCalpin) (12/08/88)

In article <7757@boring.cwi.nl> dik@cwi.nl (Dik T. Winter) writes:
>In article <159@loligo.fsu.edu> mccalpin@loligo.UUCP (John McCalpin) writes:
> > On the other hand, it is almost never necessary for users to _write_
> > is assembly language on these machines to get good performance - you
> > just need to know what the vectorizer is able to convert into efficient code
>
>When you rely on the vectorizer you may find you have bad luck (and
>especially FTN200 comes to my mind).
>-- 
>dik t. winter, cwi, amsterdam, nederland

The recent vectorizers in FTN200 (level 654 and up) are actually pretty
good.  Unfortunately, under ETA SYS V UNIX the vectorizer in the Fortran
compiler is turned OFF (!) so that you can only use the Pacific Sierra
VAST2 vectorizing preprocessor!  

My experience has also been that codes written in standard Fortran that
are designed to be efficient on the Cyber 205 or ETA-10 are also quite
fast on the Cray 1 and Cray X/MP, while the converse is generally not true.
On the Cray-2, I've not had much success at getting _anything_ to run fast,
but I do not use it often.

Of course anything that makes use hardware features of the 205/ETA (such as
bit masks) will not run well on a Cray.  This particular example is
accessible from portable Fortran (i.e. a DO loop with an IF test inside),
and runs MUCH faster on the 205 than on an X/MP.

John D. McCalpin
mccalpin@masig1.ocean.fsu.edu
mccalpin@fsu	(BITNET or MFENET)